In 2021 Intel promised to make chips not only for itself but for the whole industry — and to take on TSMC doing it. Five years, tens of billions of dollars, a departed chief executive and a government shareholder later, the factory is running. The question is who will fill it.
Easier version: read this lesson at A2/B1 level →
Ten items. Work through the four activities — say your answer out loud before you reveal it.
Three terms from the case: an anchor customer · a subsidy · an equity stake. Say the right term, then reveal.
Read each part. After it, there's a quick check (answer from memory first), then one question to discuss.
For most of its history Intel designed its own chips and made them in its own factories. By 2020 that model was under pressure: manufacturing stumbles had sent Intel’s shares down, and much of the industry now designed chips and paid Taiwan’s TSMC or South Korea’s Samsung to make them. In early 2021 Intel brought back Pat Gelsinger, an engineer who had started at the company in 1979, as chief executive.
On 23 March 2021 Gelsinger announced a strategy he called “IDM 2.0”. Intel would keep making its own chips, use outside foundries where it made sense — and become a major foundry itself, manufacturing chips for other companies from factories in the US and Europe. A new unit, Intel Foundry Services, would lead it. The first step was an estimated $20 billion for two new fabs in Chandler, Arizona. Reuters described it as a direct challenge to the only two companies that could make the most advanced chips: TSMC and Samsung.
More announcements followed. In January 2022 Intel promised more than $20 billion for two fabs in Ohio, with production planned for 2025 and room for eight fabs in total. Later that year it announced a €30 billion complex in Magdeburg, Germany, and a €5 billion plant in Wrocław, Poland. The bet was not one factory; it was a network, built ahead of demand.
1. What were the three parts of “IDM 2.0”?
2. Which new sites did Intel announce in 2021–2022?
1. Keep manufacturing its own chips, use outside foundries where useful, and become a major foundry for other companies (Intel Foundry Services).
2. Two fabs in Arizona (about $20bn), two in Ohio (more than $20bn), a €30bn complex in Magdeburg and a €5bn plant in Wrocław.
Washington wanted the same thing. In November 2024 the US Commerce Department finalised a grant of up to $7.865 billion for Intel under the CHIPS Act, to support nearly $90 billion of US investment by the end of the decade, paid out as Intel reached project milestones. But the market was less patient than the policy. Intel Foundry lost about $7 billion in 2023 and about $13.4 billion in 2024, as the company paid for new technology and new factories long before outside customers arrived in volume.
The European plans were paused in 2024. In February 2025 Intel said its first Ohio factory would now be completed only by 2030 — at least five years later than first promised. By then Gelsinger was gone. On 2 December 2024 Intel announced that he had retired, effective the day before, and named two interim co-chief executives. The chair thanked him for helping to “revitalize process manufacturing” — and said the board was committed to “restoring investor confidence”.
1. How much did the US government award Intel under the CHIPS Act, and how was it to be paid?
2. What happened to the foundry’s losses, the Ohio timeline and the chief executive?
1. Up to $7.865 billion, finalised in November 2024, paid out as Intel completed project milestones.
2. Losses grew from about $7bn (2023) to about $13.4bn (2024); the first Ohio fab slipped to 2030; Gelsinger retired in December 2024.
In March 2025 Lip-Bu Tan, a chip-industry veteran who had previously sat on Intel’s board, became chief executive. He changed the rules of the bet. In July 2025 Intel cancelled the German and Polish projects, said it would “further slow” construction in Ohio, and warned investors that it might pause or stop developing its next node, 14A, if it could not win a significant outside customer. Factories would follow customers, not the other way round.
Then came new money from unusual places. In August 2025, after President Trump had publicly said Tan should step down, the US government agreed to take a 9.9% stake in Intel: $8.9 billion for 433.3 million shares at $20.47 each, paid for with $5.7 billion of CHIPS grants not yet disbursed and $3.2 billion from a defence programme called Secure Enclave. The government also received a warrant to buy more shares under certain conditions — an incentive, Reuters reported, for Intel to keep control of its foundry. SoftBank invested $2 billion; in September Nvidia agreed to invest $5 billion and develop chips with Intel.
The engineering has started to deliver. Intel’s first PC chips on 18A came from its new Arizona fab in late 2025, and in July 2026 Intel reported revenue up 25% on the year — its strongest growth in more than fifteen years — while the foundry loss narrowed to about $2.1 billion for the quarter. But the same report still listed winning commitments from major outside customers for 14A as a key risk. The factory is real. The question is still open: will anyone else fill it?
1. What did Lip-Bu Tan change about the strategy in July 2025?
2. Describe the US government’s deal: how much, for what, and paid with what?
1. He cancelled Germany and Poland, slowed Ohio further, and made future investment (including 14A) depend on winning significant outside customers first.
2. $8.9bn for a 9.9% stake (433.3m shares at $20.47), funded by $5.7bn of undisbursed CHIPS grants and $3.2bn of Secure Enclave money, plus a conditional warrant for more shares.
Nobody in this case knows the ending. Analysts, ministers and chief executives all have to talk about it anyway — and the words they choose signal exactly how sure they are. At C1 the skill is choosing the strength that matches your evidence, and no more.
Complete each sentence with one expression: is bound to · could well · is unlikely to · is set to.
1. A new fab with no outside customers lose money in its first years — every analyst agrees.
2. The deal close next week, once the final documents are signed.
3. If a big customer signs, the 14A timetable move forward — but nobody is promising.
4. Without subsidies, a new European fab be built this decade.
1. is bound to 2. is set to 3. could well 4. is unlikely to
Complete the second sentence so it means the same as the first.
1. It is probable that Intel will need more outside investment. → Intel is need more outside investment.
2. It’s quite possible that customers will wait for 14A. → There’s a chance that customers will wait for 14A.
3. The factory is scheduled to open in 2030. → The factory is open in 2030.
4. I’m sure the share price will react. → The share price is react.
1. likely to 2. good 3. due to 4. bound to
A television presenter asks a semiconductor analyst: “Will Intel’s foundry bet pay off?” Complete the answer, then deliver it aloud in under a minute, as if live on air.
“Honestly, it’s too early to call. The losses are to continue for another year or two — that’s simply what building fabs costs. There’s a chance that 18A keeps improving, because the yields are moving in the right direction. Whether a major customer signs for 14A is the real question; one big name well change the whole story. And with the government as a shareholder, Intel is to be allowed to fail quietly.”
bound / likely · good · could / may · unlikely
Notice the shape: a near-certainty, a probability, a possibility, an improbability. A good forecast tells the listener how confident you are at every step.
Take these into the Speak tab: calibrated predictions sound credible — and protect you when the future turns out differently.
A note on the figures: amounts are rounded and come from the sources listed at the end of this tab. Intel restates some segment figures from year to year, so you may see slightly different loss numbers elsewhere. The case is still moving.
All of it out loud, against the clock. A bet that has years left to run has to be explained to people who all want a different kind of certainty.
You are advising a board that must decide whether to keep funding a long, expensive build-out. Use Intel as the case study: the old model → Gelsinger and IDM 2.0 → the network of new fabs → the CHIPS grant → the losses and the delays → Gelsinger’s departure → Tan’s new rules → the government stake, SoftBank and Nvidia → 18A and the 2026 numbers → the 14A question.
End with a forecast of what happens next, with your confidence clearly marked. Use at least five key terms (foundry, fab, process node, anchor customer, subsidy, equity stake, capacity, sunk cost, ramp up, turnaround) and at least four prediction structures of different strengths.
You are Intel’s head of foundry at an investor conference. A fund manager says: “You’ve lost tens of billions, cancelled two countries, delayed Ohio by five years, and now taxpayers own ten per cent of you. Give me one reason to believe this ends differently from every other sunk-cost story.”
Answer properly: accept what went wrong, separate the past spending from the next decision, give two pieces of evidence, and make one calibrated prediction. Then do it again in ninety seconds — the compression is the exercise.
Your company has just decided to slow the construction of a major new site until it has signed customers. Explain the decision four ways, one minute each — and notice how the strength of your predictions changes from room to room.
Prefer not to use your own organisation? Take one of these instead: